主要来源

所有链接于 2026-10-15 访问。

系统

AI Scientist-v2：2504.08066
Nature 2026：10.1038/s41586-026-10265-5
Agentic Scientist：2610.18933
ScientistTwo：2609.19644
DeepScientist：2509.26603
Jr. AI Scientist：2511.04583
FARS：2606.31651
ARIS：2605.03042
AutoResearchClaw：2605.20025
AutoSci：2605.31468
Spark-to-Paper：2608.11924
EvoScientist：2603.08127
Station：2610.08927
Agent Laboratory：2501.04227
ScientistOne：2605.26340

评测与批评

ResearchArena：2605.19156
CRUX：2607.27191（项目页）
Co-Scientist 写论文实验：2608.26701
PaperRecon：2604.01128
MLR-Bench：2505.19955
ResearchGym：2602.15112
ResearchClawBench：2606.07591
When AI Designs AI：2608.17471
AutoResearchEval：2608.14905
ResearchFlow-Eval：2610.18088
Reward hacking：2609.28614
AppliedScientist：2609.14738
SciIntegrity-Bench：2605.10246
Hidden Pitfalls：2509.08713
Paper laundering：2605.03202
Quran autoresearch：2607.18064
HPO 对比：2603.24647
Fork-and-Flush：2610.07447
引用检查工具对比：2607.22693
Agents4Science 报告：2511.15534
Science News 关于 Agents4Science 的报道
Cursor reward hacking 审计

实验室

会议

Karpathy

延伸阅读（来自反例搜索，没有逐条核实）

Science or Slop（AI 论文“科研注水”度量）：2610.00531
Xcientist（从证据到结论的可审计性）：2606.18874
自动化科研 agent 与会议“分母博弈”立场论文：2605.09915
AI 写论文与 AI 审稿的军备竞赛综述：2609.07713
ICML 2026 LLM 审稿随机对照实验：2609.19420

研究品味专题

以下为研究品味专题各页引用的外部来源（2026-10-15 访问）。arXiv 论文标题取自 Semantic Scholar 的元数据；其余条目的标签沿用页面中的链接文字。研究品味基准目录的每一行都链接到对应基准的原始来源，那些来源不在此重复列出。

实验室、评测机构与会议的报告

论文：arXiv（104 篇，按编号排序）

Inconsistency in Conference Peer Review: Revisiting the 2014 NeurIPS Experiment（arXiv 2109.09774）
How do authors’ perceptions of their papers compare with co-authors’ perceptions and peer-review decisions?（arXiv 2211.12966）
Has the Machine Learning Review Process Become More Arbitrary as the Field Has Grown? The NeurIPS 2021 Consistency Experiment（arXiv 2306.03262）
Understanding the Effects of RLHF on LLM Generalisation and Diversity（arXiv 2310.06452）
Large language models surpass human experts in predicting neuroscience results（arXiv 2403.03230）
Large Language Models Reflect Human Citation Patterns with a Heightened Citation Bias（arXiv 2405.15739）
Generating Interesting Scientific Ideas using Knowledge Graphs and LLMs: Evaluations with 100 Research Group Leaders（arXiv 2405.17044）
Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers（arXiv 2409.04109）
Human-LLM Compound System for Scientific Ideation through Facet Recombination and Novelty Evaluation（arXiv 2409.14634）
Human Creativity in the Age of LLMs: Randomized Experiments on Divergent and Convergent Thinking（arXiv 2410.03703）
CycleResearcher: Improving Automated Research via Automated Review（arXiv 2411.00816）
IdeaBench: Benchmarking Large Language Models for Research Idea Generation（arXiv 2411.02429）
Evaluating LLMs’Divergent Thinking Capabilities for Scientific Idea Generation with Minimal Context（arXiv 2412.17596）
Evaluating Sakana’s AI Scientist: Bold Claims, Mixed Results, and a Promising Future?（arXiv 2502.14297）
All That Glitters is Not Novel: Plagiarism in AI Generated Research（arXiv 2502.16487）
Accelerating scientific discovery with Co-Scientist（arXiv 2502.18864）
ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decomposition（arXiv 2503.21248）
The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search（arXiv 2504.08066）
Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?（arXiv 2504.13837）
Predicting Empirical AI Research Outcomes with Language Models（arXiv 2506.00794）
AlphaEvolve: A coding agent for scientific and algorithmic discovery（arXiv 2506.13131）
The Ideation-Execution Gap: Execution Outcomes of LLM-Generated versus Human Research Ideas（arXiv 2506.20803）
Will Compute Bottlenecks Prevent an Intelligence Explosion?（arXiv 2507.23181）
ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution（arXiv 2509.19349）
DeepScientist: Advancing Frontier-Pushing Scientific Findings Progressively（arXiv 2509.26603）
Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity（arXiv 2510.01171）
LLM-REVal: Can We Trust LLM Reviewers Yet?（arXiv 2510.12367）
BadScientist: Can a Research Agent Write Convincing but Unsound Papers that Fool LLM Reviewers?（arXiv 2510.18003）
Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)（arXiv 2510.22954）
Kosmos: An AI Scientist for Autonomous Discovery（arXiv 2511.02824）
Mathematical exploration and discovery at scale（arXiv 2511.02864）
Jr. AI Scientist and Its Risk Report: Autonomous Scientific Exploration from a Baseline Paper（arXiv 2511.04583）
AlphaResearch: Accelerating New Algorithm Discovery with Language Models（arXiv 2511.08522）
When AI Does Science: Evaluating the Autonomous AI Scientist KOSMOS in Radiation Biology（arXiv 2511.13825）
Exploring the use of AI authors and reviewers at Agents4Science（arXiv 2511.15534）
What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity（arXiv 2511.15593）
Even with AI, Bijection Discovery is Still Hard: The Opportunities and Challenges of OpenEvolve for Novel Bijection Construction（arXiv 2511.20987）
Training AI Co-Scientists Using Rubric Rewards（arXiv 2512.23707）
Why LLMs Aren’t Scientists Yet: Lessons from Four Autonomous Research Attempts（arXiv 2601.03315）
Global Optimization for Combinatorial Geometry Problems Revisited in the Era of LLMs（arXiv 2601.05943）
Who Owns Creativity and Who Does the Work? Trade-offs in LLM-Supported Research Ideation（arXiv 2601.12152）
Towards Execution-Grounded Automated AI Research（arXiv 2601.14525）
Learning to Discover at Test Time（arXiv 2601.16175）
Semi-Autonomous Mathematics Discovery with Gemini: A Case Study on the Erdős Problems（arXiv 2601.22401）
AI-Assisted Scientific Assessment: A Case Study on Climate Change（arXiv 2602.09723）
Towards Autonomous Mathematics Research（arXiv 2602.10177）
Is this Idea Novel? An Automated Benchmark for Judgment of Research Ideas（arXiv 2603.10303）
AI Can Learn Scientific Taste（arXiv 2603.14473）
HindSight: Evaluating LLM-Generated Research Ideas via Future Impact（arXiv 2603.15164）
LLMs learn scientific taste from institutional traces across the social sciences（arXiv 2603.16659）
GIANTS: Generative Insight Anticipation from Scientific Literature（arXiv 2604.09793）
SciPredict: Can LLMs Predict the Outcomes of Scientific Experiments in Natural Sciences?（arXiv 2604.10718）
Debate as Reward: A Multi-Agent Reward System for Scientific Ideation via RL Post-Training（arXiv 2604.16723）
When AI reviews science: Can we trust the referee?（arXiv 2604.23593）
Agentic AI Scientists Are Not Built For Autonomous Scientific Discovery（arXiv 2605.08956）
LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search（arXiv 2605.09764）
AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration（arXiv 2605.20025）
Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation（arXiv 2605.21491）
Scientific reasoning does not reliably translate into scientific forecasting in frontier AI（arXiv 2605.22681）
AI Research Agents Narrow Scientific Exploration（arXiv 2605.27905）
SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?（arXiv 2605.30329）
Contemporary AI lacks the imagination to diverge or negate in science（arXiv 2606.08251）
On the Limits of LLM-as-Judge for Scientific Novelty Assessment（arXiv 2606.12071）
From AGI to ASI（arXiv 2606.12683）
First Proof Second Batch（arXiv 2606.18119）
Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty（arXiv 2606.25198）
Measuring the Gap Between Human and LLM Research Ideas（arXiv 2607.01233）
ResearchStudio-Idea: An Evidence-Grounded Research-Ideation Skill Suite from ML Conference Outcomes（arXiv 2607.04439）
Human diversity fuels collective creativity that large language models cannot simulate or sustain（arXiv 2607.26899）
Can AI agents conduct open-ended AI research? Early evidence from two case studies（arXiv 2607.27191）
Diversifying Personalized Research Ideation against AI-Induced Homogenization（arXiv 2607.28087）
Style Wins, Substance Loses: A Diagnosis of LLM-as-Judge in Idea Generation（arXiv 2608.01666）
Long-Horizon Autonomous Architecture Research with a Language-Model Agent: A Behavioural Case Study（arXiv 2608.01995）
Adversarial Fast-Moving Real-World Domains as Test Beds for Benchmarking AI Scientist Capabilities（arXiv 2608.03569）
An AI Scientist that Doesn’t Drift: Taste, Structure, and Falsifiable Findings in a Quadruped Navigation Research Loop（arXiv 2608.07542）
Beyond Final Scores: A Systematic Evaluation of Agents for Long-Horizon AI Research and Development（arXiv 2608.13417）
AI Research Preference Models（arXiv 2608.13940）
A Human Audit of OpenAIs AI-Generated Mathematical Proofs（arXiv 2608.14673）
Personalized Auto-Research: Towards a True AI Co-Scientist（arXiv 2608.14881）
Historical Backtesting for Scientific Question Discovery: A Protocol and Astronomy Pilot（arXiv 2608.16795）
The Problem Is the Problem: Towards Scalable Mathematical Discovery（arXiv 2608.16977）
When AI Designs AI: Innovation or Imitation?（arXiv 2608.17471）
Ideation Arena: Evaluating LLM Generated Research Ideas with Battle-style Human Expert Assessment（arXiv 2608.29696）
Can LLM Agents Discover? Evaluating Creativity on ML Engineering Tasks（arXiv 2608.30047）
Can Large Language Models Forecast What Researchers Study Next?（arXiv 2609.00747）
Creators have difficulty abandoning ideas they generated（arXiv 2609.05115）
Agentic ML Exploration (A-MLE) for Ads Ranking（arXiv 2609.08248）
Comment on arXiv:2607.01233: Survivorship Bias in Published-Paper Baselines for Research-Idea Distributions（arXiv 2609.15996）
The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks（arXiv 2609.25804）
Learning to Ideate for Scientific Impact（arXiv 2609.29802）
RINI: Seeing the Prior Is Not Enough（arXiv 2609.33284）
How Strong Is the Evidence for the Artificial Hivemind? Reevaluating Evidence for the Open-Ended Homogeneity of Language Models（arXiv 2609.33936）
Reinforcing Agentic Creativity in Scientific Ideation with Night Science（arXiv 2609.35706）
Old Ideas, Novel Problems: The Instability of LLM-Based Novelty Evaluation（arXiv 2610.02022）
Are We Measuring Scientific Intelligence? Rethinking the Evaluation of AI Scientists（arXiv 2610.04915）
Priced Guidance: Can Language Models Generate Future Research Ideas?（arXiv 2610.04976）
TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts（arXiv 2610.06824）
Fork-and-Flush: Escaping Idea Basins in Autoresearch Agents（arXiv 2610.07447）
Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station（arXiv 2610.08927）
Verify Less, Evolve More: Training Idea-Level Critics for Verification-Efficient ML Evolving Agents（arXiv 2610.08993）
DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck（arXiv 2610.17184）
Large Language Models Extend Known Research But Do Not Create New Research Directions（arXiv 2610.17439）
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery（arXiv 2610.18933）
Learning Individual Scientific Taste with Privileged Hindsight Distillation（arXiv 2610.19719）

主要来源

研究品味专题

实验室、评测机构与会议的报告

论文：arXiv（104 篇，按编号排序）

论文：期刊、OpenReview、NBER 等

文章、博客与访谈